GPT-5.5 and Gemma 4 Launch While Google Chrome Silently Installs Local AI Models
New Large Language Model Releases
GPT-5.5 Instant Rollout: OpenAI has begun rolling out the GPT-5.5 Instant model within ChatGPT, featuring significant enhancements to its chatting capabilities. The new model is reportedly twice as expensive to access via API compared to the previous GPT-5.4 version.
Google Releases Gemma 4 with Multi-Token Prediction: Google has launched the Gemma 4 MTP model series, which utilizes Multi-Token Prediction technology to improve performance. The release includes multiple variants ranging from 2B to 31B parameters, optimized for assistant-style tasks.
- Gemma 4 MTP released
- https://blog.google/innovation-and-ai/technology/developers-tools/multi-token-prediction-gemma-4
- https://huggingface.co/google/gemma-4-31B-it-assistant
- https://huggingface.co/google/gemma-4-26B-A4B-it-assistant
- https://huggingface.co/google/gemma-4-E4B-it-assistant
- https://huggingface.co/google/gemma-4-E2B-it-assistant
Model Performance & Benchmarking
Major Updates to LLM Debate Leaderboard: The LLM Debate Benchmark has added a suite of new models, including GPT-5.5, Grok 4.3, and DeepSeek V4 Pro, to its adversarial testing suite. Despite the influx of new high-reasoning models from various providers, Opus 4.7 continues to hold the top position on the leaderboard.
Claude Series Intelligence and Cost Analysis: A new analysis of the Claude AI family illustrates the correlation between intelligence benchmarks and token costs across the Haiku, Sonnet, and Opus tiers. The data specifically highlights the efficiency gains of Claude 3.5 Sonnet compared to earlier models in the ecosystem.
Mistral Model Coding Benchmarks: Recent testing of Mistral’s newest models on the Terminal Bench 2.0 evaluation platform reveals performance variations across the "Small" and "Medium" iterations. The results indicate that the Mistral Medium 3.5 High model offers a significant improvement in win rates for code-related tasks.
Industry & Product News
Google Chrome Silent Local AI Installation: Users have discovered that Google Chrome is automatically downloading a 4 GB local AI model onto devices without seeking explicit user consent. This discovery has prompted concerns regarding storage management, privacy, and the environmental impact of forced background downloads.